Query and Index Statistics Collector

track_activities

Parameter description: Collects statistics about the commands that are being executed in session.

Type: SUSET

Value range: Boolean

• on indicates that the statistics collection function is enabled.
• off indicates that the statistics collection function is disabled.

Default value: on

track_counts

Parameter description: Collects statistics about data activities.

Type: SUSET

Value range: Boolean

• on indicates that the statistics collection function is enabled.
• off indicates that the statistics collection function is disabled.

When the database to be cleaned up is selected from the AutoVacuum automatic cleanup process, the database statistics are required. In this case, the default value is set to on.

Default value: on

track_io_timing

Parameter description: Collects statistics about I/O invoking timing in the database. The I/O timing statistics can be queried by using the pg_stat_database parameter.

Type: SUSET

Value range: Boolean

• If this parameter is set to on, the collection function is enabled. In this case, the collector repeatedly queries the OS at the current time. As a result, large numbers of costs may occur on some platforms. Therefore, the default value is set to off.
• off indicates that the statistics collection function is disabled.

Default value: off

track_functions

Parameter description: Collects statistics about invoking times and duration in a function.

Type: SUSET

When the SQL functions are set to inline functions queried by the invoking, these SQL functions cannot be traced no matter these functions are set or not.

Value range: enumerated values

• pl indicates that only procedural language functions are traced.
• all indicates that SQL and C language functions are traced.
• none indicates that the function tracing function is disabled.

Default value: none

track_activity_query_size

Parameter description: Specifies byte counts for each active session's currently executing command. This parameter is used for pg_stat_activity.query and

pgxc_stat_activity.query columns. If no unit is specified, the unit is byte by default.

Type: POSTMASTER

Value range: an integer ranging from 100 to 102400

Default value: 1024

update_process_title

Parameter description: Collects statistics updated with a process name each time the server receives a new SQL statement.

The process name can be viewed on Windows task manager by running the ps command.

Type: SUSET

Value range: Boolean

• on indicates that the statistics collection function is enabled.
• off indicates that the statistics collection function is disabled.

Default value: off

track_thread_wait_status_interval

Parameter description: Specifies the interval of collecting the thread status information periodically.

Type: SUSET

Value range: an integer ranging from 0 to 1440, in minutes.

Default value: 30min

enable_save_datachanged_timestamp

Parameter description: Specifies whether to record the time when INSERT, UPDATE, DELETE, or EXCHANGE/TRUNCATE/DROP PARTITION is performed on table data.

Type: USERSET

Value range: Boolean

• on indicates that the time when an operation is performed on table data will be recorded.
• off indicates that the time when an operation is performed on table data will not be recorded.

Default value: on

enable_save_dataaccess_timestamp

Parameter description: Specifies whether to record the last access time of a table. This parameter is supported only by 8.2.1.210 and later cluster versions.

Type: USERSET

Value range: Boolean

• on indicates that the last access time of the table is recorded.
• off indicates that the last access time of the table is not recorded.

Default value: off

instr_unique_sql_count

Parameter description: Specifies whether to collect Unique SQL statements and the maximum number allowed.

Type: SIGHUP

Value range: an integer ranging from 0 to INT_MAX

• If it is set to 0, Unique SQL statistics are not collected.
• If the value is greater than 0, the number of Unique SQL statements collected on the CN cannot exceed the value of this parameter. When the number of collected Unique SQL statements reaches the upper limit, the collection is stopped. In this case, you can increase the value of reload to continue the collection.

Default value: 0

If a new value is smaller than the original value, the Unique SQL statistics collected on the CN will be cleared. Note that the clearing operation is performed by the background thread of the resource management module. If the GUC parameter use_workload_manager is set to off, the clearing operation may fail. In this case, you can use the reset_instr_unique_sql function for clearing.

instr_unique_sql_timeout

Parameter description: Specifies the lifetime of a Unique SQL statement. The background thread of StatCollector checks all Unique SQL statements every hour. If a Unique SQL statement is not executed for more than instr_unique_sql_timeout hours, the Unique SQL statement will be deleted. This feature is supported in 8.1.2 or later.

Type: SIGHUP

Value range: an integer ranging from 0 to INT_MAX, in hours

• The value 0 indicates that expired Unique SQL statements will not be deleted.
• If the value is greater than 0, the Unique SQL statement that is not executed for more than instr_unique_sql_timeout hours will be deleted.

Default value: 24

track_sql_count

Parameter description: Specifies whether to collect statistics on the number of the SELECT, INSERT, UPDATE, DELETE, and MERGE INTO statements that are being executed in each session, the response time of the SELECT, INSERT, UPDATE, and DELETE statements, and the number of DDL, DML, and DCL statements.

Type: SUSET

Value range: Boolean

• on indicates that the statistics collection function is enabled.
• off indicates that the statistics collection function is disabled.

Default value: on

• The track_sql_count parameter is restricted by the track_activities parameter.
  • If track_activities is set to on and track_sql_count is set to off, a warning message indicating that track_sql_count is disabled will be displayed when the view gs_sql_count, pgxc_sql_count, gs_workload_sql_count, pgxc_workload_sql_count, global_workload_sql_count, gs_workload_sql_elapse_time, pgxc_workload_sql_elapse_time, or global_workload_sql_elapse_time are queried.
  • If both track_activities and track_sql_count are set to off, two logs indicating that track_activities is disabled and track_sql_count is disabled will be displayed when the views are queried.
  • If track_activities is set to off and track_sql_count is set to on, a log indicating that track_activities is disabled will be displayed when the views are queried.

• If this parameter is disabled, querying the view returns 0.

enable_track_wait_event

Parameter description: Specifies whether to collect statistics on waiting events, including the number of occurrence times, number of failures, duration, maximum waiting time, minimum waiting time, and average waiting time.

Type: SIGHUP

Value range: Boolean

• on indicates that the statistics collection function is enabled.
• off indicates that the statistics collection function is disabled.

Default value: off

• The enable_track_wait_event parameter is restricted by track_activities. Its functions cannot take effect no matter whether it is enabled if track_activities is disabled.
• When track_activities or enable_track_wait_event is disabled, if you query the get_instr_wait_event function, gs_wait_events view, or pgxc_wait_events view, a message is displayed indicating that the GUC parameter is disabled and the query result is 0.
• If track_activities or enable_track_wait_event is disabled during cluster running, GaussDB(DWS) will not collect statistics on waiting events. However, statistics that have been collected are not affected.

enable_wdr_snapshot

Parameter description: Specifies whether to enable the performance view snapshot function. After this function is enabled, GaussDB(DWS) will periodically create snapshots for some system performance views and save them permanently. In addition, it will accept manual snapshot creation requests.

Type: SIGHUP

Value range: Boolean

• on indicates that the snapshot function is enabled.
• off indicates that the snapshot function is disabled.

Default value: off

• If the create_wdr_snapshot function is executed to manually create a view when the enable_wdr_snapshot parameter is disabled, a message is displayed indicating that the GUC parameter is not enabled.
• If the enable_wdr_snapshot parameter is modified during the snapshot creation process, the snapshot that is being created is not affected. The modification takes effect when the snapshot is manually or periodically created next time.

wdr_snapshot_interval

Parameter description: Specifies the interval for automatically creating performance view snapshots.

Type: SIGHUP

Value range: an integer ranging from 10 to 180, in minutes

Default value: 60

• The value of this parameter must be set in accordance with the cluster load. You are advised to set this parameter to a value greater than the time required for creating a snapshot.
• If the value of wdr_snapshot_interval is less than the time required for creating a snapshot, the system will skip this snapshot creation because it finds that the previous snapshot creation is not complete when the time for this automatic snapshot creation arrives.

wdr_snapshot_retention_days

Parameter description: Specifies the maximum number of days for storing performance snapshot data.

Type: SIGHUP

Value range: an integer ranging from 1 to 15, in days

Default value: 8

• If enable_wdr_snapshot is enabled, snapshot data that has been stored for wdr_snapshot_retention_days days will be automatically deleted.
• The value of this parameter must be set in accordance with the available disk space. A larger value requires more disk space.
• The modification of this parameter does not take effect immediately. The expired snapshot data will be cleared only when a snapshot is automatically created next time.

enable_parallel_analyze

Parameter description: Specifies whether to use parallel sampling for internal and foreign table analysis. This parameter is supported only by clusters of version 9.1.0 or later.

Type: USERSET

Value range: Boolean

• true indicates that parallel sampling is used for internal and foreign table analysis.
• false indicates that parallel sampling is not used for internal and foreign table analysis.

Default value: true

• When enable_parallel_analyze is set to true and analyzing foreign tables, try to avoid adding NOT NULL constraints to the target foreign table columns to prevent constraint failure due to data source changes. Currently, parallel sampling does not support materialized views. If analyze fails due to such reasons, set this parameter to false.
• Currently, parallel sampling only supports analyzing ordinary column-store internal tables. This optimization does not take effect when the internal table uses hstore/hstore_opt or is declared as a replicated table.
• Currently, parallel sampling only supports analyzing foreign tables stored in parquet/orc format. This optimization does not take effect when the foreign table is in another format.

parallel_analyze_workers

Parameter description: Specifies the number of concurrent threads for parallel analyze sampling. This parameter is supported only by clusters of version 9.1.0 or later.

Type: USERSET

Value range: an integer ranging from 0 to 64

Default value: 10

The value of this parameter should correspond to the cluster load. When the cluster load is low, you can increase the parameter value appropriately based on the cluster configuration to further improve the efficiency of analyze execution.

analyze_sample_multiplier

Parameter description: Specifies the multiplier for the stripe sampling rate used in analyzing foreign tables. This parameter is supported only by clusters of version 9.1.0 or later.

Type: SUSET

Value range: an integer ranging from 0 to 100. 0 indicates that the stripe sampling rate is 100%.

Default value: 3